$63 Billion: The Largest AI Infrastructure Financing in History! Broadcom CEO Gets Personally Involved, Wall Street Closely Watching Massive Debt Issuance in Q1 Next Year
This deal to help Anthropic lease chips almost fell through due to "indigestion" in the AI debt market. Broadcom CEO Hock Tan urgently facilitated a rare collaboration between banks and Blackstone by offering "higher interest rates plus residual value guarantees" as a concession. Broadcom acted both as the seller and guarantor, creating a controversial “circular financing” loop that deeply ties the company to Anthropic. The issuance of approximately $21 billion in high-grade bonds next quarter will serve as a real stress test.
A record-breaking AI infrastructure financing deal nearly fell through at the last moment.
According to The Wall Street Journal on October 27, Broadcom CEO Hock Tan personally stepped in for emergency negotiations with executives of several banks during the final weekend, ultimately securing an AI chip financing transaction with a total scale of up to $63 billion—making it the largest artificial intelligence infrastructure debt financing in the world to date.
Codenamed "Project North Star," the transaction aims to help AI startup Anthropic lease custom chips developed jointly by Broadcom and Google. It was disclosed in parallel with Anthropic's IPO filing. The deal is jointly led by Bank of America, Citigroup, and Morgan Stanley, with Blackstone committing $9 billion as anchor capital for another portion of the financing.
Reports indicate that the completion of this deal was far from smooth. According to informed sources, in recent weeks, some participating banks considered withdrawing due to rising interest rates and increasing AI industry risks, feeling that the offered compensation was insufficient to cover the risks. Hock Tan then launched a "last-minute" negotiation over the weekend, agreeing to raise loan interest rates and provide stronger loss mitigation assurances to the banks, thereby stabilizing the situation.
Additionally, according to Bloomberg, the banks plan to issue about $21 billion in bonds to the U.S. investment-grade corporate bond market as early as Q1 2027 to hedge their approximately $42 billion loan exposure. This issuance plan is being closely watched by Wall Street and Silicon Valley as a key indicator of investor appetite for AI project financing.
The Largest AI Financing in History: Rare Alliance Between Banks and Private Credit
Beyond its scale, the "North Star" deal set another record: Banks and private credit institutions—two typically competing sources of capital—joined forces for the first time to complete an AI infrastructure financing of this magnitude.
According to The Wall Street Journal, the entire $63 billion financing consists of two parts:
About $42 billion in senior secured loans, underwritten by a syndicate of banks, with Broadcom providing guarantees;
Another approximately $18 billion in higher-risk subordinate debt, not backed by Broadcom’s guarantee, with half provided by Blackstone.
Banks plan to distribute about half of their loans (around $22 billion) into the investment-grade bond market. According to Bloomberg, the final bond issuance will be about $21 billion with an expected five-year maturity, with principal repaid in installments during the term. The specific issuance size and timing will depend on market conditions.
This is also the second major chip financing completed by Broadcom this year. Previously, Broadcom, in conjunction with Apollo Global Management and Blackstone, closed a $35 billion transaction.
Broadcom’s Multifaceted Role: Supplier, Guarantor, Negotiator
In the "North Star" deal, Broadcom played the highly unusual role of wearing multiple hats.
As a chip supplier, Broadcom partnered with Google to develop the tensor processing units (TPUs) required by Anthropic. As a financing intermediary, Broadcom essentially provided "vendor financing"—the deal is structured so that a special purpose vehicle (SPV) borrows $63 billion to purchase the chips, then leases them to Anthropic, with Anthropic’s rental income used to repay the debt.
Broadcom’s key contribution lies in its balance sheet endorsement. As reported by Bloomberg, Broadcom disclosed in this week’s regulatory filings that its maximum potential guarantee liability is: in the event of Anthropic default, Broadcom must cover the difference between “72.5% of the outstanding lease amount” and “the value realized from disposing of the AI rack assets.” Broadcom considers the likelihood of this guarantee being triggered to be low. This “residual value support” mechanism is precisely why banks were able to include about $42 billion of senior secured loans in their underwriting.
Hock Tan’s direct involvement in the negotiations further underscores the strategic significance of this deal for Broadcom. According to The Wall Street Journal, he communicated with bank executives one by one during the final weekend, ultimately securing commitments in exchange for higher interest rates.
Critics have labeled the financing model represented by "Project North Star" as “circular financing.”
In Anthropic’s IPO prospectus, the company revealed it expects to pay more than $125.2 billion for leasing Broadcom chips, with Broadcom and Google respectively promising to provide $42 billion and $20 billion of funding support to Anthropic. This means Broadcom is simultaneously the chip vendor, a provider of financing, and a guarantor—creating a closed-loop flow of funds between the parties.
"North Star" is a sequel to the earlier Anthropic-Broadcom chip financing deal codenamed "Big Sky." According to The Wall Street Journal, "Big Sky" was financed by Apollo, though Apollo has not yet completed distributing some of the debt to investors. An Apollo spokesperson stated that the company has "completed the syndication of all 'Big Sky' debt."
AI Debt Market Struggles With “Indigestion,” Banks Under Pressure
The dramatic negotiations around “North Star” reflect deeper market pressures: The supply of AI-related debt has already shown signs of “indigestion” in the market.
According to The Wall Street Journal, since the beginning of this year, hundreds of billions of dollars in AI-related bonds and loans have flooded the market. Many banks are still digesting previously committed loans, limiting additional capacity. Some banks have been forced to sell loans at a discount to free up capital, while others choose to hold on, further narrowing their ability to participate in new deals.
Bloomberg reports that global AI-related financing has exceeded $455 billion so far this year, but asset managers' willingness to continue expanding AI debt exposure is waning.
Morgan Stanley’s situation is particularly illustrative. As quoted by The Wall Street Journal citing debt fund managers, Morgan Stanley is struggling to issue about $300 million in asset-backed securities for Project Blitz, a Kansas City data center project operated by AI compute startup Lambda. Despite doubling the bond yields, there has been little interest from buyers.
Interest rate trends are another sword of Damocles hanging over this transaction. As reported by The Wall Street Journal, banks plan to wait until next year’s chip deliveries before selling bonds to fund managers. If rates rise sharply by then, the banks could face losses from having to sell at a discount—because they have already committed to an upper limit on loan interest rates. Some banks may then opt to hold loans rather than sell at a loss.
Wall Street Awaits Q1 “Mega Debt Issuance” Next Year
The ultimate completion of “North Star” is both a reflection of the ever-increasing scale of AI infrastructure financing and a signal that this market is approaching its limits.
For Wall Street, the issuance of approximately $21 billion in high-grade bonds in the first quarter of next year will be a real-world stress test—measuring just how much appetite investors still have in the context of an AI financing boom and mounting market digestion pressure. For Broadcom, the deal consolidates its position as an “indispensable” hub in the AI supply chain, but also further binds its balance sheet to Anthropic's fate.
As The Wall Street Journal notes, Broadcom’s roles as supplier, guarantor, and negotiator in this deal “reflect the intertwined nature of the big bets being placed on AI, which are reshaping the economic landscape and involve high stakes for all participants.”
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
Outstanding performance, record shareholder returns—will SK Hynix counterattack short sellers?
JPMorgan believes that SK Hynix's revenue and operating profit for the third quarter both reached record highs, exceeding market consensus by approximately 7% and 9% respectively. Its 100 trillion Korean won shareholder return plan is about 2.5 times the consensus. JPMorgan maintains its overweight rating with a target price of 2.7 million Korean won, expecting that the better-than-expected results combined with the return plan will shift market sentiment on the stock from negative to positive.

Morgan Stanley Quantitative Team: AI Chip Stocks Haven’t Bottomed Yet, Recommend “Reducing Positions on Rebounds”
Morgan Stanley’s quant team has warned that the sell-off in semiconductor and data center stocks is not over yet, as the current trend of the chip index resembles that of the internet bubble in 2000, indicating continued medium-term downside risk. They recommend taking advantage of rebounds to reduce positions, increase allocation to low-volatility and value stocks, and position in AI application companies. The report points out that the AI theme is shifting from upstream infrastructure to downstream applications, with AI adoption companies demonstrating relative resilience.

JPMorgan: If AI-driven profits persist, US stocks can withstand a 10-year Treasury yield rising to 5.5%-6.0%
JPMorgan pointed out that based on approximately 80 years of historical data, if the current forecast for forward EPS growth exceeding 20% is achieved, the S&P 500 has the capacity to absorb a rise in the 10-year U.S. Treasury yield to the 5.5%-6.0% range—especially under the premise that AI-driven productivity growth supports a high valuation environment.


